Private Consensus using Chaotic Oscillator-Based Encryption

Abstract

Distributed average consensus is a fundamental feature of multi-agents systems; yet, in several cases agents are reluctant to disclose their initial conditions, e.g., due to their sensitivity about private data. Consequently, ensuring the privacy of such information against honest but curious neighbors becomes a mandatory necessity. In this paper we propose to implement a privacy-preserving consensus strategy that exploits, for this purpose, unpredictable chaotic phenomena, such as the trend of variables in a Chua oscillator. The initial conditions are then split into two fragments, one of which always remains hidden in the node, while the other is exchanged after undergoing oscillator-dependent manipulation, adding an extra layer of security to what is exchanged over the network. In this way, the combination of the two fragments converges to the average of the true initial conditions of each node. The paper is complemented by a simulation campaign aimed at numerically demonstrating the effectiveness of the proposed approach.